diff --git a/src/calibre/ebooks/rtf/input.py b/src/calibre/ebooks/rtf/input.py
index 55f42ae4d5..ff20793f39 100644
--- a/src/calibre/ebooks/rtf/input.py
+++ b/src/calibre/ebooks/rtf/input.py
@@ -169,6 +169,21 @@ class RTFInput(InputFormatPlugin):
         with open('styles.css', 'ab') as f:
             f.write(css)
 
+    def preprocess(self, fname):
+        self.log('\tPreprocessing to convert unicode characters')
+        try:
+            data = open(fname, 'rb').read()
+            from calibre.ebooks.rtf.preprocess import RtfTokenizer, RtfTokenParser
+            tokenizer = RtfTokenizer(data)
+            tokens = RtfTokenParser(tokenizer.tokens)
+            data = tokens.toRTF()
+            fname = 'preprocessed.rtf'
+            with open(fname, 'wb') as f:
+                f.write(data)
+        except:
+            self.log.exception(
+            'Failed to preprocess RTF to convert unicode sequences, ignoring...')
+        return fname
 
     def convert(self, stream, options, file_ext, log,
                 accelerators):
@@ -177,8 +192,9 @@ class RTFInput(InputFormatPlugin):
         from calibre.ebooks.rtf2xml.ParseRtf import RtfInvalidCodeException
         self.log = log
         self.log('Converting RTF to XML...')
+        fname = self.preprocess(stream.name)
         try:
-            xml = self.generate_xml(stream.name)
+            xml = self.generate_xml(fname)
         except RtfInvalidCodeException:
             raise ValueError(_('This RTF file has a feature calibre does not '
             'support. Convert it to HTML first and then try it.'))
diff --git a/src/calibre/ebooks/rtf/preprocess.py b/src/calibre/ebooks/rtf/preprocess.py
new file mode 100644
index 0000000000..07e6d41fac
--- /dev/null
+++ b/src/calibre/ebooks/rtf/preprocess.py
@@ -0,0 +1,344 @@
+#!/usr/bin/env python
+# vim:fileencoding=UTF-8:ts=4:sw=4:sta:et:sts=4:ai
+from __future__ import with_statement
+
+__license__   = 'GPL v3'
+__copyright__ = '2010, Gerendi Sandor Attila'
+__docformat__ = 'restructuredtext en'
+
+"""
+RTF tokenizer and token parser. v.1.0 (1/17/2010)
+Author: Gerendi Sandor Attila
+
+At this point this will tokenize a RTF file then rebuild it from the tokens.
+In the process the UTF8 tokens are altered to be supported by the RTF2XML and also remain RTF specification compilant.
+"""
+
+class tokenDelimitatorStart():
+    def __init__(self):
+        pass
+    def toRTF(self):
+        return b'{'
+    def __repr__(self):
+        return '{'
+
+class tokenDelimitatorEnd():
+    def __init__(self):
+        pass
+    def toRTF(self):
+        return b'}'
+    def __repr__(self):
+        return '}'
+
+class tokenControlWord():
+    def __init__(self, name, separator = ''):
+        self.name = name
+        self.separator = separator
+    def toRTF(self):
+        return self.name + self.separator
+    def __repr__(self):
+        return self.name + self.separator
+
+class tokenControlWordWithNumericArgument():
+    def __init__(self, name, argument, separator = ''):
+        self.name = name
+        self.argument = argument
+        self.separator = separator
+    def toRTF(self):
+        return self.name + repr(self.argument) + self.separator
+    def __repr__(self):
+        return self.name + repr(self.argument) + self.separator
+
+class tokenControlSymbol():
+    def __init__(self, name):
+        self.name = name
+    def toRTF(self):
+        return self.name
+    def __repr__(self):
+        return self.name
+
+class tokenData():
+    def __init__(self, data):
+        self.data = data
+    def toRTF(self):
+        return self.data
+    def __repr__(self):
+        return self.data
+
+class tokenBinN():
+    def __init__(self, data, separator = ''):
+        self.data = data
+        self.separator = separator
+    def toRTF(self):
+        return "\\bin" + repr(len(self.data)) + self.separator + self.data
+    def __repr__(self):
+        return "\\bin" + repr(len(self.data)) + self.separator + self.data
+
+class token8bitChar():
+    def __init__(self, data):
+        self.data = data
+    def toRTF(self):
+        return "\\'" + self.data
+    def __repr__(self):
+        return "\\'" + self.data
+
+class tokenUnicode():
+    def __init__(self, data, separator = '', current_ucn = 1, eqList = []):
+        self.data = data
+        self.separator = separator
+        self.current_ucn = current_ucn
+        self.eqList = eqList
+    def toRTF(self):
+        result = '\\u' + repr(self.data) + ' '
+        ucn = self.current_ucn
+        if len(self.eqList) < ucn:
+            ucn = len(self.eqList)
+            result =  tokenControlWordWithNumericArgument('\\uc', ucn).toRTF() + result
+        i = 0
+        for eq in self.eqList:
+            if i >= ucn:
+                break
+            result = result + eq.toRTF()
+        return result
+    def __repr__(self):
+        return '\\u' + repr(self.data)
+
+
+def isAsciiLetter(value):
+    return ((value >= 'a') and (value <= 'z')) or ((value >= 'A') and (value <= 'Z'))
+
+def isDigit(value):
+    return (value >= '0') and (value <= '9')
+
+def isChar(value, char):
+    return value == char
+
+def isString(buffer, string):
+    return buffer == string
+
+
+class RtfTokenParser():
+    def __init__(self, tokens):
+        self.tokens = tokens
+        self.process()
+        self.processUnicode()
+
+    def process(self):
+        i = 0
+        newTokens = []
+        while i < len(self.tokens):
+            if isinstance(self.tokens[i], tokenControlSymbol):
+                if isString(self.tokens[i].name, "\\'"):
+                    i = i + 1
+                    if not isinstance(self.tokens[i], tokenData):
+                        raise BaseException('Error: token8bitChar without data.')
+                    if len(self.tokens[i].data) < 2:
+                        raise BaseException('Error: token8bitChar without data.')
+                    newTokens.append(token8bitChar(self.tokens[i].data[0:2]))
+                    if len(self.tokens[i].data) > 2:
+                        newTokens.append(tokenData(self.tokens[i].data[2:]))
+                    i = i + 1
+                    continue
+
+            newTokens.append(self.tokens[i])
+            i = i + 1
+
+        self.tokens = list(newTokens)
+
+    def processUnicode(self):
+        i = 0
+        newTokens = []
+        ucNbStack = [1]
+        while i < len(self.tokens):
+            if isinstance(self.tokens[i], tokenDelimitatorStart):
+                ucNbStack.append(ucNbStack[len(ucNbStack) - 1])
+                newTokens.append(self.tokens[i])
+                i = i + 1
+                continue
+            if isinstance(self.tokens[i], tokenDelimitatorEnd):
+                ucNbStack.pop()
+                newTokens.append(self.tokens[i])
+                i = i + 1
+                continue
+            if isinstance(self.tokens[i], tokenControlWordWithNumericArgument):
+                if isString(self.tokens[i].name, '\\uc'):
+                    ucNbStack[len(ucNbStack) - 1] = self.tokens[i].argument
+                    newTokens.append(self.tokens[i])
+                    i = i + 1
+                    continue
+                if isString(self.tokens[i].name, '\\u'):
+                    x = i
+                    j = 0
+                    i = i + 1
+                    replace = []
+                    partialData = None
+                    ucn = ucNbStack[len(ucNbStack) - 1]
+                    while (i < len(self.tokens)) and (j < ucn):
+                        if isinstance(self.tokens[i], tokenDelimitatorStart):
+                            break
+                        if isinstance(self.tokens[i], tokenDelimitatorEnd):
+                            break
+                        if isinstance(self.tokens[i], tokenData):
+                            if len(self.tokens[i].data) >= ucn - j:
+                                replace.append(tokenData(self.tokens[i].data[0 : ucn - j]))
+                                if len(self.tokens[i].data) > ucn - j:
+                                    partialData = tokenData(self.tokens[i].data[ucn - j:])
+                                i = i + 1
+                                break
+                            else:
+                                replace.append(self.tokens[i])
+                                j = j + len(self.tokens[i].data)
+                                i = i + 1
+                                continue
+                        if isinstance(self.tokens[i], token8bitChar) or isinstance(self.tokens[i], tokenBinN):
+                            replace.append(self.tokens[i])
+                            i = i + 1
+                            j = j + 1
+                            continue
+                        raise BaseException('Error: incorect utf replacement.')
+
+                    #calibre rtf2xml does not support utfreplace
+                    replace = []
+
+                    newTokens.append(tokenUnicode(self.tokens[x].argument, self.tokens[x].separator, ucNbStack[len(ucNbStack) - 1], replace))
+                    if partialData != None:
+                        newTokens.append(partialData)
+                    continue
+
+            newTokens.append(self.tokens[i])
+            i = i + 1
+
+        self.tokens = list(newTokens)
+
+
+    def toRTF(self):
+        result = []
+        for token in self.tokens:
+            result.append(token.toRTF())
+        return "".join(result)
+
+
+class RtfTokenizer():
+    def __init__(self, rtfData):
+        self.rtfData = []
+        self.tokens = []
+        self.rtfData = rtfData
+        self.tokenize()
+
+    def tokenize(self):
+        i = 0
+        lastDataStart = -1
+        while i < len(self.rtfData):
+
+            if isChar(self.rtfData[i], '{'):
+                if lastDataStart > -1:
+                    self.tokens.append(tokenData(self.rtfData[lastDataStart : i]))
+                    lastDataStart = -1
+                self.tokens.append(tokenDelimitatorStart())
+                i = i + 1
+                continue
+
+            if isChar(self.rtfData[i], '}'):
+                if lastDataStart > -1:
+                    self.tokens.append(tokenData(self.rtfData[lastDataStart : i]))
+                    lastDataStart = -1
+                self.tokens.append(tokenDelimitatorEnd())
+                i = i + 1
+                continue
+
+            if isChar(self.rtfData[i], '\\'):
+                if i + 1 >= len(self.rtfData):
+                    raise BaseException('Error: Control character found at the end of the document.')
+
+                if lastDataStart > -1:
+                    self.tokens.append(tokenData(self.rtfData[lastDataStart : i]))
+                    lastDataStart = -1
+
+                tokenStart = i
+                i = i + 1
+
+                #Control Words
+                if isAsciiLetter(self.rtfData[i]):
+                    #consume <ASCII Letter Sequence>
+                    consumed = False
+                    while i < len(self.rtfData):
+                        if not isAsciiLetter(self.rtfData[i]):
+                            tokenEnd = i
+                            consumed = True
+                            break
+                        i = i + 1
+
+                    if not consumed:
+                        raise BaseException('Error (at:%d): Control Word without end.'%(tokenStart))
+
+                    #we have numeric argument before delimiter
+                    if isChar(self.rtfData[i], '-') or isDigit(self.rtfData[i]):
+                        #consume the numeric argument
+                        consumed = False
+                        l = 0
+                        while i < len(self.rtfData):
+                            if not isDigit(self.rtfData[i]):
+                                consumed = True
+                                break
+                            l = l + 1
+                            i = i + 1
+                            if l > 10 :
+                                raise BaseException('Error (at:%d): Too many digits in control word numeric argument.'%[tokenStart])
+
+                        if not consumed:
+                            raise BaseException('Error (at:%d): Control Word without numeric argument end.'%[tokenStart])
+
+                    separator = ''
+                    if isChar(self.rtfData[i], ' '):
+                        separator = ' '
+
+                    controlWord = self.rtfData[tokenStart: tokenEnd]
+                    if tokenEnd < i:
+                        value = int(self.rtfData[tokenEnd: i])
+                        if isString(controlWord, "\\bin"):
+                            i = i + value
+                            self.tokens.append(tokenBinN(self.rtfData[tokenStart:i], separator))
+                        else:
+                            self.tokens.append(tokenControlWordWithNumericArgument(controlWord, value, separator))
+                    else:
+                        self.tokens.append(tokenControlWord(controlWord, separator))
+                    #space delimiter, we should discard it
+                    if self.rtfData[i] == ' ':
+                        i = i + 1
+
+                #Control Symbol
+                else:
+                    self.tokens.append(tokenControlSymbol(self.rtfData[tokenStart : i + 1]))
+                    i = i + 1
+                continue
+
+            if lastDataStart < 0:
+                lastDataStart = i
+            i = i + 1
+
+    def toRTF(self):
+        result = []
+        for token in self.tokens:
+            result.append(token.toRTF())
+        return "".join(result)
+
+
+if __name__ == "__main__":
+    import sys
+    if len(sys.argv) < 2:
+        print ("Usage %prog rtfFileToConvert")
+        sys.exit()
+    f = open(sys.argv[1], 'rb')
+    data = f.read()
+    f.close()
+
+    tokenizer = RtfTokenizer(data)
+    parsedTokens = RtfTokenParser(tokenizer.tokens)
+
+    data = parsedTokens.toRTF()
+
+    f = open(sys.argv[1], 'w')
+    f.write(data)
+    f.close()
+
+